BACKGROUND: Considerable evidence has shown that the use of computational algorithms to combine pretreatment clinical and pathology results can enhance predictions of patient outcome. The aim of this study was to prove that the application of such methods to predict patient-specific likelihoods of organ-confined (OC) prostate carcinoma (PCA) may be helpful to patients and physicians when they are choosing an optimal treatment for carcinoma of the prostate. METHODS: The authors used clinical and quantitative pathology results from the biopsy specimens of 817 PCA patients who had been evaluated at a large national pathology reference laboratory. The pathology parameters that were measured included the number of positive cores, Gleason grades and score, percentage of tumor involvement, and the tumor location. The pathologic stage of these cases, as determined by results from radical prostatectomy, lymphadenectomy, or bone scan, categorized the PCA as either OC, non-OC due to capsular penetration only (NOC-CP) or advanced disease with metastasis (NOC-Mets), i.e., seminal vesicle and/or lymph-node positive or bone-scan positive. There were a total of 481 OC cases, 185 NOC-CP cases, and 151 NOC-Mets cases. Patient-specific prediction models were trained by ordinal logistic regression (OLOGIT) and genetically engineered neural networks (GENNs), and the resulting trained models were validated by biopsy information from an independent set of 116 PCA patients. RESULTS: When the authors applied a cutoff of >or= 35% for the n = 817 training set of OC, NOC-CP, and NOC-Mets predictive probabilities, the OLOGIT model predicted OC PCA with an accuracy of 91%, whereas the GENN model predicted the same with an accuracy of 95%. When the authors employed the n = 116 validation set (76 OCs, 31 NOC-CPs, and 9 NOC-Mets), the OLOGIT and GENN models correctly identified OC PCA with 91% and 97% accuracy, respectively. CONCLUSIONS: The value of combining patient pretreatment diagnostic pathology parameters to make predictions concerning the postoperative extent of pathology was illustrated clearly in this study. This finding further confirms the need to pursue such approaches for PCA disease management in the future, especially with the increasing prevalence of clinical T1c (American Joint Committee on Cancer, 1977) disease. Copyright 2001 American Cancer Society.
BACKGROUND: Considerable evidence has shown that the use of computational algorithms to combine pretreatment clinical and pathology results can enhance predictions of patient outcome. The aim of this study was to prove that the application of such methods to predict patient-specific likelihoods of organ-confined (OC) prostate carcinoma (PCA) may be helpful to patients and physicians when they are choosing an optimal treatment for carcinoma of the prostate. METHODS: The authors used clinical and quantitative pathology results from the biopsy specimens of 817 PCA patients who had been evaluated at a large national pathology reference laboratory. The pathology parameters that were measured included the number of positive cores, Gleason grades and score, percentage of tumor involvement, and the tumor location. The pathologic stage of these cases, as determined by results from radical prostatectomy, lymphadenectomy, or bone scan, categorized the PCA as either OC, non-OC due to capsular penetration only (NOC-CP) or advanced disease with metastasis (NOC-Mets), i.e., seminal vesicle and/or lymph-node positive or bone-scan positive. There were a total of 481 OC cases, 185 NOC-CP cases, and 151 NOC-Mets cases. Patient-specific prediction models were trained by ordinal logistic regression (OLOGIT) and genetically engineered neural networks (GENNs), and the resulting trained models were validated by biopsy information from an independent set of 116 PCA patients. RESULTS: When the authors applied a cutoff of >or= 35% for the n = 817 training set of OC, NOC-CP, and NOC-Mets predictive probabilities, the OLOGIT model predicted OC PCA with an accuracy of 91%, whereas the GENN model predicted the same with an accuracy of 95%. When the authors employed the n = 116 validation set (76 OCs, 31 NOC-CPs, and 9 NOC-Mets), the OLOGIT and GENN models correctly identified OC PCA with 91% and 97% accuracy, respectively. CONCLUSIONS: The value of combining patient pretreatment diagnostic pathology parameters to make predictions concerning the postoperative extent of pathology was illustrated clearly in this study. This finding further confirms the need to pursue such approaches for PCA disease management in the future, especially with the increasing prevalence of clinical T1c (American Joint Committee on Cancer, 1977) disease. Copyright 2001 American Cancer Society.
Authors: Ying Huang; Sumit Isharwal; Alexander Haese; Felix K H Chun; Danil V Makarov; Ziding Feng; Misop Han; Elizabeth Humphreys; Jonathan I Epstein; Alan W Partin; Robert W Veltri Journal: BJU Int Date: 2010-09-28 Impact factor: 5.588
Authors: Clement G Yedjou; Solange S Tchounwou; Richard A Aló; Rashid Elhag; BereKet Mochona; Lekan Latinwo Journal: Int J Sci Acad Res Date: 2021-10-30
Authors: Shahrokh F Shariat; Michael W Kattan; Andrew J Vickers; Pierre I Karakiewicz; Peter T Scardino Journal: Future Oncol Date: 2009-12 Impact factor: 3.404
Authors: Catherine Wilson; Pamela J Maxwell; Daniel B Longley; Richard H Wilson; Patrick G Johnston; David J J Waugh Journal: PLoS One Date: 2012-05-09 Impact factor: 3.240